A monitoring approach for runtime service discovery

摘要

Effective runtime service discovery requires identification of services based on different service characteristics such as structural, behavioural, quality, and contextual characteristics. However, current service registries guarantee services described in terms of structural and sometimes quality characteristics and, therefore, it is not always possible to assume that services in them will have all the characteristics required for effective service discovery. In this paper, we describe a monitor-based runtime service discovery framework called MoRSeD. The framework supports service discovery in both push and pull modes of query execution. The push mode of query execution is performed in parallel to the execution of a service-based system, in a proactive way. Both types of queries are specified in a query language called SerDiQueL that allows the representation of structural, behavioral, quality, and contextual conditions of services to be identified. The framework uses a monitor component to verify if behavioral and contextual conditions in the queries can be satisfied by services, based on translations of these conditions into properties represented in event calculus, and verification of the satisfiability of these properties against services. The monitor is also used to support identification that services participating in a service-based system are unavailable, and identification of changes in the behavioral and contextual characteristics of the services. A prototype implementation of the framework has been developed. The framework has been evaluated in terms of comparison of its performance when using and when not using the monitor component.